Intel Hala Point unveiled as the world's largest neuromorphic system

[Redaktion]

With 1.15 billion neurons, the Intel Hala Point is the world's largest neuromorphic system. Taking advantage of the Loihi 2 processor and deployed at Sandia National Laboratories, this system's capacity is similar to that of an owl brain and will be used to solve computing problems in device physics as well as computer science.

https://www.notebookcheck.net/Intel-Hala-Point-unveiled-as-the-world-s-largest-neuromorphic-system.828895.0.html

[Fadeev]

It is also suitable for LLM: language tokens ( 102 billion Russian ) and the vectors are sitting in Knowledge Graphs. The question is tokenized, i.e. it turns into a vector. Next, a set (for example, 1000 pieces) of similar vectors by Kulback statistics is selected from the language tokens according to the question vector. Then, the same statistics compare the vector of the question and 1000 binary vectors with the result - up to 10 vectors with their conversion into an answer. The length of the binary vector is 300 bits and processing is quite capable of Hala Point with its 8-bit commands. It won't even be too long - no more than 1 second. A possible negative is the creeping of AI.

[NikoB]

The nonsense of all such marketing self-praise is that no one really knows 100% how neurons and synapses work at their full potential. They only think that they understand part of the processes in the most advanced biological neuro-machine known to us - the brain. I am 100% sure that they will not be able to create intelligence superior to human (in 100% of aspects) in the next 100 years. In some narrow analytical areas it is possible, but at the same creative level it is impossible. Even with all the resources of the planet. The technology is too primitive to achieve this even with a consumption of 1000000 more.

In addition, the goal of real AI (unlike the extremely forgetful human brain) is to remember all the facts down to the smallest detail, and not just associations. This is what will allow artificial intelligence to surpass the biological neuro-machine, because he will operate with accurate facts, and not with fog in his head, like people (even the smartest ones, if you take something other than what they are aimed at in a fairly short time)..

It will take at least another 50-70 years (and then only if humanity does not end up in a massive global slaughter with all the ensuing consequences for science and progress) before scientists will possibly discover a 100% picture of how the human brain works. And probably even this period is extremely optimistic.